The association between community mental health nursing and hospital admissions for people with serious mental illness: a systematic review

This systematic review aimed to investigate the association between exposure to community mental health nursing care and admission to hospital (a suitable proxy for relapse) [20] in people with serious mental illness.

Observational studies (including case-control, cohort and cross-sectional studies) and clinical trials (including non-randomised controlled trials and randomised controlled trials) [i.e. study design] examining the effect of care provided by mental health nurses [i.e. exposure] to community-dwelling patients with SMI (i.e. schizophrenia, bipolar disorder and major depression) [i.e. participants] were eligible for inclusion in this review. The intervention could be compared with any other model of care [i.e. comparator]. For this review, we defined a community mental health or psychiatric nurse as a person holding a formal specialised qualification in psychiatric/mental health nursing and had been registered, credentialed or licenced to practise in that capacity (e.g. registered mental health/psychiatric nurse) and was working primarily in the community. It is important to note that while the review protocol did not explicitly mention ‘community’ mental health nurses, it was implied. Studies evaluating multidisciplinary team-based models of care, specific mental health nurse administered clinical interventions (e.g. family work, cognitive behavioural therapy) or care provided in secure or other inpatient settings were excluded. No restriction was applied to the language or date of publication.

The MEDLINE [Ovid, 1946 to present] search strategy is presented in Table 1. This strategy was adapted, as necessary, for the following databases: CINAHL [EBSCOHost, 1937 to present], PubMed [NCBI, 1966 to present], EMBASE [Ovid, 1947 to present], Nursing & Allied Health Database [ProQuest, inception to present], Health Source: Nursing/Academic Edition [EBSCOHost, inception to present], PsycINFO [Ovid, 1806 to present], Ovid Nursing [Ovid, 1946 to present], ProQuest Dissertations and Theses Global [ProQuest, 1743 to present], The Cochrane Library [1992 to present], and Web of Science [Clarivate Analytics, 1975 to present]. All databases were searched from the date of inception to June 2017. The reference lists of included publications, and articles citing the included publications, were also hand searched to identify potentially eligible studies. Clinical trial registries (i.e. clinicaltrials.​gov; WHO Clinical Trials) were searched to identify any on-going or unpublished trials. The search strategy was initially implemented in July 2017 and updated in November 2019.

Publications identified through the searches were exported into reference management software (EndNote X8, Clarivate Analytics, Boston, USA). Duplicate records were excluded. The reference management file was subsequently exported to systematic review software (Covidence, Veritas Health Innovation, Melbourne, Australia) for screening. Titles and abstracts of all identified publications were screened for eligibility against the review selection criteria (i.e. eligible study design, participants, exposure and outcomes) by two reviewers (a process shared by all reviewers), independently. Remaining publications underwent full-text screening by two reviewers (a process shared by all reviewers), independently. At both stages of the process, disagreements were arbitrated by a third reviewer.

The primary outcome of interest was hospital admission (i.e. admission/readmission to a psychiatric inpatient facility), which served as a proxy for SMI relapse. This outcome was selected because it is associated with a deterioration in health and social exclusion; it is also the primary focus of community psychiatric services, is widely used in mental health services research and is economically meaningful [21]. Secondary outcomes were hospital length of stay, emergency department presentations, crisis team referral, duration of crisis team treatment, crisis house referral, detention under mental health law and adverse events.

Data from eligible publications were extracted using a customised data extraction tool. The tool collected information on study characteristics, research methodology, participant characteristics, interventions, comparators, outcomes, results and new references. Data extraction was performed by three reviewers (ML, MJ, RG), independently, with disagreements adjudicated by a fourth reviewer (AE).

The risk of bias of observational studies was assessed using the Risk Of Bias In Non-randomised Studies of Interventions (ROBINS-I) tool [22]. The ROBINS-I tool determines the risk of bias across seven distinct domains, including baseline and time-varying confounding, intervention classification, co-interventions, participant selection, outcome measurement, missing data and selective reporting bias. Two reviewers (ML, RG) independently evaluated the risk of bias and rated studies as having a low, moderate, severe, critical or unclear risk of bias. A third reviewer (MJ) was consulted if there was disagreement between reviewers.

The risk of bias of clinical trials was evaluated using the Cochrane Collaboration risk of bias tool [23]. The Cochrane tool assesses risk across seven domains, including allocation concealment, sequence generation, blinding of participants, personnel and outcome assessors, selective outcome reporting, incomplete outcome data and other sources of bias. Two reviewers (ML, RG) independently rated the risk of bias for each item as low, unclear or high risk. A third reviewer (MJ) was invited to arbitrate when consensus was not reached between reviewers.

Given the considerable methodological heterogeneity of studies and that no single comparable outcome was reported in more than three studies, results could not be combined by means of meta-analysis. The results were instead presented using narrative synthesis. This synthesis was undertaken by one reviewer (ML) and cross-checked by two others (RG, MJ).

This review was prospectively registered with PROSPERO [CRD42017058694]. A detailed summary of the review methods is set out in the protocol, which has been published elsewhere [24]. The review is also reported in accordance with PRISMA reporting guidelines for systematic reviews [25].

The search identified 1153 publications, of which 258 duplicates were removed (Fig. 1). During title/abstract screening, 860 publications were excluded. At full-text screening, a further 32 publications were excluded, mostly because they reported the wrong intervention (N = 24). Three publications, reporting two discrete studies, met the selection criteria and were included in the review.

Both included studies were conducted in community settings in England (Table 2). One was an observational study [26], the other a randomised controlled trial [27, 28]. Duration of follow-up was 18 months [27, 28] and 3 years [26].

The 2 studies involved a total of 356 patients with serious mental illness. McCrone et al. [27]/Muijen et al. [28] restricted their sample to patients with psychotic disorders, whilst Barr et al. [26] included all patients with a SMI. Participants were primarily middle-aged, and just over half were male (gender was not reported in the Barr et al. [26] study).

Both studies reported community mental health nursing as the intervention under investigation. However, there were differences between the two interventions. Barr et al. [26] examined the association between community psychiatric nurse (CPN) contact and psychiatric admission versus no CPN contact. McCrone et al. [27]/Muijen et al. [28] compared the effect of intensive CPN support, as measured by increased CPN contact time, versus traditional CPN support. Patients in the intensive support team received twice the amount of contact with CPNs than patients in the traditional CPN group. Table 2 provides additional, albeit limited details (due to insufficient reporting in the publications), of the interventions used in this study.

The two studies reported distinct outcomes, two of which were relevant for this review: admission to hospital [26] and health service use (i.e. hospital admissions and emergency department presentations) [27, 28].

The methodological quality of Barr et al. [26] was appraised using ROBINS-I (Table 3). The study was found to be of moderate methodological quality, with six of the seven parameters having a moderate risk of bias. Of note, was the high risk of confounding, as the authors made no adjustments for confounders in their analysis. The Cochrane risk of bias tool was used to appraise the McCrone et al. [27]/Muijen et al. [28] trial, which was found to be of low methodological quality (Table 3). The study received an uncertain rating for four (allocation concealment, sequence generation, blinding of participants, blinding of assessors) of the seven parameters on the Cochrane risk of bias tool. Both studies were more than a decade old, with Barr et al. [26] published 18 years ago and McCrone et al. [27]/Muijen et al. [28] published 24 years ago. Accordingly, these studies predate the introduction of quality standards for the reporting of clinical trials.

The focus of this review was on hospital readmission of patients with SMI. The review did not include studies that focused on quality of life, recovery, reduced suicide risk, vocational needs or overall satisfaction with care—all of which are areas where community mental health nursing may have played an important role. While admission to hospital was used as a proxy measure for relapse of SMI, this may not be an accurate measure of relapse since home treatment or crisis teams have been part of standard care for people in mental health crisis in the UK since 2000. In addition, people may be admitted to psychiatric inpatient facilities for clozapine initiation, which may not represent a relapse of SMI. However, admission to hospital has been used as a measure of relapse in other studies [35] and is an objective outcome used in psychiatric services. The inclusion of crisis teams as a proxy measure for relapse might yield different results in future reviews, although the view of the authors is that generating new primary data should be the priority for future research.